The acquisition policy uses a two-stage selection mechanism: Stage 1 probabilistically samples a Rank Factor Class weighted by freshness, while Stage 2 applies Thompson sampling, Upper Confidence Bound, or epsilon-greedy selection across individual claim posteriors within that class.